Census population estimates 2011-2019

[jimrtex]

Just for fun, this is what would happen if they non-retroactively reinstated the no states can lose a seat rule for 2020:

It increases the House size by 18.

Wouldn't this lead to extreme increases in the size of the House if repeated in the future?

[jimrtex]

Probably not so extreme as one thinks, and the House chamber could accommodate 600 based on discussions I had when visiting there 20 years ago.

Iowa has gone from 11 to 4, North Dakota from 3 to 1, and Vermont from 6 to 1.

In general the increase would be proportional to the rate of gain of the US minus the growth rate of the slowest growing (fastest declining) state. So perhaps 45 to 50 per decade.

If the proposed rule had first been applied in 2010, what would the size of the House been in 2010, and what would it be in 2020 if the procedure had been retained?

[jimrtex]

Here's my annual projection from the new estimates. I used the July 2018 estimates and the April 2010 Census base to get an annual growth rate. This correctly accounts for the 8 and a quarter year period between the Census and the estimate. I then applied the annual growth rate to the 2010 reapportionment population to get the 2020 projection. This accounts for the extra overseas population used in reapportionment but not for redistricting. Ten years is a long stretch for a simple model like this, but here are the projected changes.

AL -1
AZ +1
CA -1
CO +1
FL +2
IL -1
MI -1
MN -1
MT +1
NY -1
NC +1
OH -1
OR +1
PA -1
RI -1
TX +3
WV -1

Compared to last year, CA loses one and MT gains one. That is what my model based on the two years from 2015-2017 projected, so it appears that that short-term trend has continued enough to influence the whole decade. The bubble seats in this projection are based on the last five awarded and the next five in line.
The last five awarded are IL-17, FL-29, TX-39, NY-26, and MT-2 (#435).
The next five in line are CA53, AL-7, MN-8, OH-15, and VA-12.

The alternate projection based on just the prior two years of estimates to determine the rate of growth matches the full decade projection now.

Based on resident population, projected exponentially for the last 1.75 years, I have California narrowly keeping its 53rd seat, and Montana staying at one seat. But an additional 1000 population for Montana or -30K for California would swap the two.

The exponential projection is quite optimistic for California. Growth in California was quite flat for the first half of the decade through 2015 at 327K per year. The last three years have been 256K, 190K, and 158K. That is, the second derivative has been negative. The exponential growth projection would add 506K over the next 1.75 years. Taking 7/8 of the 2016-2018 increase would only add 305K. Not only is California not going to meet the exponential projection it will miss it badly.

We will likely miss the California Paradox where California's share of population is increasing, while its share of representation is decreasing. Currently, California is slightly above the USA growth rate (projected at 7.3% for the decade), but will likely fall below the national rate by 2020.

So this will give Montana the 435th seat. Montana's growth is not robust, but will increase by about 9.0% during the decade. Montana's growth rate is volatile, subject to energy prices. Expansion of the Williston Basin into Montana depends on sustained higher oil prices, along with a way to transport the oil. Regina is the closest population center to Williston, but Billings is the closest domestic center. Both Billings and Bismarck benefit from rentals to the Williston Basin.

A negative for Montana is it's COLD. But Montana is not going to drop much below its current rate of growth in the next 1.75 years. It could be on the cusp between one and two seats for decades to come. After losing the 2nd seat at the 1990 census, Montana has been just short of gaining it back - at times missing simply because of the timing of the census.

The only realistic contender for the 435th seat is Minnesota. But while its growth has increased a bit the last few years, it should still be short of holding on to its 8th district. Exponentially projected growth would give Minnesota 68K more persons, about 32K short of the needed population. Linear growth based on 2016-2018 would net 77K. This would still leave Minnesota 23K short.  It would really need to pop its growth for the next two year, and the increase for 2018 was actually less than 2017.


Realized changes for 2018, were for Arizona to gain its 10th district, and Rhode Island to lose its 2nd. Montana surpassed Rhode Island in population during the year.

Quite likely changes by 2020 are for Florida to gain its 28th and Texas 39th, at the expense of Alabama to 6 and Ohio to 15.

The final change if it happens is Montana 2nd at the expense of California to 52.

[jimrtex]

I used a linear growth rate estimation based on changes across the whole decade.

If that's the case you should add the overseas military personnel times the same growth rate to the state populations since the 2020 Census will. MT has about double the participation rate of CA in overseas military - 0.5% compared to 0.24% for CA. When I took the military population out of my projection CA was back on top using the whole decade. However, even without the military adjustment MT still wins using a short term 2 or 3 year average for projection.

Most of the overseas population is military personnel and their dependents.

The Census Bureau's rationalization is that military personnel are not living overseas by choice. The real reason is that the Census can easily get records from other federal agencies.

The US government requires that overseas civilian citizens be able to vote in their state of last residence. It is pretty hypocritical to require that they be able to vote for representatives, while not including them in the determination of the number of representatives they may vote for.

The US should maintain voter registration for civilian overseas and all military personnel (this would avoid providing information about location of military personnel to states). Overseas voters should be able to vote at embassies, consulates, other population centers, and military bases.

An overseas voter could go to a federal voting center, prove his identity. The voting center would contact the state of (last) residence and generate a ballot. The voter would vote, and the results would be sent back to the state.

With these registration records, the Census Bureau could contact civilians living overseas and determine the number of associated dependents.

In 2010, about 1,000,000 of 1,048,000 were associated with the military (95%+). About 400,000 were military and 600,000 were dependents.

Since 2010 overseas military has declined by about half. OTOH, there has been little decline in numbers stationed in countries that are suitable for dependents: Japan, Germany, Korea, Italy, UK, Spain and Belgium (countries with more than 1K military personnel).

The DOD administrative records include 3 potential states: home of record - which is generally where someone resided prior to (re)enlistment; legal address - where someone pays state taxes (tends to be heavy on Texas, Florida, Washington, Alaska, Wyoming, and Tennessee which have no income taxes, but have military bases which can be used to establish legal residency); and last duty session.

If available, the Census Bureau uses the home of record, perhaps reasoning that was the last place that was under control of the service member - and likely to produce less distortion of the apportionment population.

If we assume that the distribution of home states has not changed, and that the overseas military has declined by 50%, but not in locations where dependents can live. That means we could take (800/1050)/1.073  which would be ratio of overseas population to resident population in 2020 to that in 2010.

Then for each state take the 2010 ratio of overseas/resident population and multiply by this factor and multiply this by the 2020 projected resident population.

This assumes that factors that lead to enlistment do not vary over time, but the number who do enlist is proportional to the resident population of a state.

[jimrtex]

Imagine how physically tiny NV-01 is going to be in the next set of maps.

It is kind of like the two districts in the Northern Territory of Australia. One does not get out of Darwin.

[jimrtex]

19 states are gaining ground relative to the USA:

Western states (all but AK, HI, NM, and WY): AZ, CA, CO, ID, MT, NV, OR, UT, WA
Central tier: ND, SD, TX
South Atlantic (DE to FL, except MD): DE, FL, GA, NC, SC, VA
Tennessee: TN

These states can gain representation. The other 32 states are losing ground relative to the USA and can lose representation.

The following table is based on exponentially projected 2020 resident population. Above the line it shows the relative quotient for 2020 gainers: AZ(10), CO(8), FL(29), NC(14), OR(6), and TX(39). They will have notched these gains by 2020, and will be moving away from the line. The other states are growing slower than the national average and are at risk of losing additional seats. States like MA and NE are just below the national average in growth, and are potentially decades away from change. It is more likely that there will be a change in trends.

Pretty certain losses for 2030 are CT, IL, NJ, NY(2), PA, and WI. Possible losers are IL(2), IN, MI, and OH. These are based on comparing margin for 2020 and growth rates from 2010 to 2020 relative to the national growth rate of 7.3.

States below the line that lost a seat are AL(6), IL(17), MI(13), MN(7), NY(26), OH(15), PA(17), RI(1), WV(2). These states have notched a loss for 2020, and will be dropping further. With the exception of MN, they won't be making up lost ground.

Other states are growing faster than the national average and have the possibility of gaining seats. Fairly certain gainers for 2030 are FL(2), ID, TX(3), UT, and WA. Possible gainers are AZ, CO, FL(3), GA, NV, TX(4), VA.

Because there are more possible gainers than possible losers, there is a greater chance that the potential losses will occur (more eager contenders to grab the seat away from the sluggards). But it is likely that all the potential gainers will not be realized (too much competition).

Special cases are CA, MN, and MT, which have growth rates near to the national average AND are close to the threshold. They can ping pong back and forth. The MN growth rate is increasing a bit, which could turn it into a gainer by 2030, while CA growth rate is slowing substantially which could turn it into a loser. It likely will already lose its 53rd due to the slowdown, and could lose additional seats. California loses/gains seats for roughly every 2% difference from the national rate. A 5.3% increase vs. a national rate of 7.3% would cost another seat. The MT economy is too resource dependent, and has no major cities to sustain growth (compared to say Boise, Des Moines, or Omaha).


NY 23 1.092
PA 15 1.085
MO  7 1.078
MA  8 1.078
IL 15 1.076 ?
KY  5 1.076
MD  7 1.068
OH 14 1.060 ?
MI 12 1.051 ?
NY 24 1.047 x
CT  4 1.047 x
NE  2 1.044
IN  8 1.040 ?
NC 14 1.027(+)
CO  8 1.024(+)
WI  7 1.023 x
OR  6 1.022(+)
PA 16 1.019 x
NJ 11 1.019 x
CA 51 1.015
AZ 10 1.015(+)
IL 16 1.011 x
NY 25 1.006 x
FL 29 1.006(++)
TX 39 1.005(+++)
CA 52 0.9999
--------------------------------------------------
MT  2 0.9992
MN  7 0.994(-)
AL  6 0.993(-)
OH 15 0.991(-)
VA 12 0.984 ?
CA 54 0.981
RI  1 0.980(-)
TX 40 0.979 x
MI 13 0.973(-)
FL 30 0.971 x
NY 26 0.968(-)
GA 15 0.968 ?
WA 11 0.964 x
ID  3 0.961 x
PA 17 0.961(-)
WV  2 0.961(-)
TX 41 0.955 x
UT  5 0.953 x
IL 17 0.953(-)
TN 10 0.948
NC 15 0.946
FL 31 0.940 x
TX 42 0.932
AZ 11 0.918 ?
NV  4 0.911 ?
DE  2 0.910
TX 43 0.910 ?
FL 32 0.910 ?
SC  8 0.908
VA 13 0.905
GA 16 0.905
CO  9 0.903 ?

States that are losing ground relative to the USA, but are at least 10% above losing a seat.
AR(4), HI(2), IA(4), KS(4), LA(6), ME(2), MS(4), NM(3), OK(5). They won't lose the seat by 2030.

States that are losing ground relative to the USA, and already have the minimum number of seats.

AK(1),  VT(1), and WY(1).

States that are gaining ground relative to the USA, but are at least 10% below gaining a seat, and won't gain a seat by 2030 (barring oil going above $100 per barrel).

ND(1) and SD(1)

[jimrtex]

So question, was New Jersey adjusted down? In 2017, the population was past 9 million but this 2018 Census states the population is 8.9 million while at the same time showing NJ with population growth of 0.22% from last year. Not that it matters all that much since the twelfth district is safe in 2020 but its just kind of bothering me.

https://www.census.gov/quickfacts/fact/table/nj/IPE120217

Census estimates are based on estimates of demographic changes due to births, deaths, in-migration, and out-migration.

Birth and death estimates are based on birth and death data from the National Heath Sevice (NHS). This lags the estimate date by 1-1/2 years (for the July 1, 2018 estimate is based on data for calendar year 2016). The 2018 estimate is based on a projection of the earlier data.

The estimate made for 2017 did not have 2016 data, so the death and birth data was based on projection from earlier data. But in 2018, we can correct the 2017 estimate. Even with accurate birth and death records, place of residence may be misleading. A widow from Florida who moved to live with he daughter in North Carolina, before she died, will show up as a North Carolinian dying, and a Floridian getting a year older (since no death or move was recorded). Some mothers move after giving birth. If their husband was in Afghanistan, she may have move to be nearer her parents home during the pregnancy, rather than being alone near some military base.

Domestic migration is estimated based on IRS, Medicare, and SSA records - which can apparently be anonymized enough such that you can find where a taxpayer has changed residence, and the number of exemptions and perhaps the age. Since this does not cover those who don't file, they project this to the total population This could be fairly current, since most taxpayers have filed by July 1. It got really complicated for international migration.

METHODOLOGY FOR THE UNITED STATES POPULATION ESTIMATES: VINTAGE 2018 (PDF)

I think they try to track the number of persons from the 2010 Census:

Let's say there are N 47 YO white males in 2010.

X% die,
Y% move to another state.
Z% move to another country.
W% move from another country.

Then for 2011 they have an estimated number of 48 YO white males, etc.

To try to make sense of the estimates, I compared the 2018 vintage annual components of change with the 2017 vintage annual components of change for New York.

The changes in birth and deaths appear just to be noise. But there was a dramatic decrease in the net inflow of international migration, which goes back most of the decade. That is, the Census Bureau now believes that they had been overestimating net international migration. This is close to a 200K difference over the decade.

I don't know the source of the error. Perhaps they were using the ACS and could not find the immigrants. Perhaps there have been people leaving the US. These are hard to detect, because the US government doesn't care if you leave. They, in theory, care if you enter the country.

Of course it is is possible that the ACS does a particularly poor job of counting persons who are illegally in the country, and avoid anything they believe might result in detection.

If there has been a change in methodology with respect to international migration, this would have a particularly profound effect on the estimates for New York and California.

[jimrtex]

Puerto Rico lost 4% of its population last year (a record) and that was not only because of out-migration (-123K people), but also because a -7K death surplus. A loss of 130K people in total.

Since the Census 2000, more than 1 million Puerto Ricans have left the island - or about 1/4 of the population ...

It was more a 7K birth deficit. Annual births have declined from 41K to 24K during the decade.

I suspect those who leave may be disproportionately in child-bearing ages. Yep ...

If we compare numbers in 5-year age groups from 2012 to 2017, when they would have advanced to the next age group.

20-25 in 2012, 25-30 in 2017 84.9% *** groups with largest share of births
25-30 in 2012, 30-34 in 2017 83.8% ***
30-34 in 2012, 35-39 in 2017 87.4%
35-39 in 2013, 40-44 in 2017 90.2%
...
55-59 in 2013, 60-64 in 2017 93.1%

At older ages, you see a sharp decline, but that is because of people dying before they age 5 years. This may have a small effect on those in their 50s.

There is a strong male-female differential.

25-29 in 2012, 30-34 in 2017, male 18.5% left, female 14.1% left.

In a normal population, males outnumber females until in 40s to 50s, when earlier male deaths balance the population. But here you have potential mates leaving the island.

If we look at children, there is also difference.

0-4 in 2012, 5-9 in 2017 86.9%
5-10 in 2012, 10-14 in 2017 89.3%
10-14 in 2012, 15-19 in 2017 89.2%
15-19 in 2012, 20-24 in 2017 86.6%

The youngest are more likely to be living in non-established families, and it is easier to move if you don't have an established career, a house, and the children in school. You likely have family on the mainland, whom you can use for at least temporary housing, and may know of jobs.

Older children, who are in school, and may have more difficulty in transitioning to English-based schools.  Migration matches that of those 30 years older (their parents).

The oldest group which were teenagers in 2012, has now entered adulthood and are beginning to migrate.

[jimrtex]

Yes, New Jersey's was adjusted down, just like New York's was adjusted down by like a quarter million.

Looking at the components of change: birth, deaths, domestic migration, and international migration, and comparing the 2017 and 2018 vintage estimates for 2017, there was a dramatic decrease in the estimate of net international migration between 2010-2017 from 7.2 million to 6.7 million. Remember that net includes both immigration and emigration. The changes for births, deaths, and domestic migration were minimal.

The changes in international migration were extremely concentrated, with New York (-265K), California (-147K) and New Jersey (-121K) suffering the most.

MD -29
NV -23
FL -14
GA -14
CT -13
LA -13
PA -12
NC -11

There were a few gainers:

TX +36
AZ +33
MI +18
WA +18
OH +11

I suspect that the Census Bureau has refined their estimates of people leaving the US, which would reasonably be the hardest to measure, since they no longer exist in the US. Residents for census purposes include temporary residents such as those attending college or on long-term work assignments, It also includes persons who move to the USA to work and travel, with no intent to permanently reside in the USA.

Some people change their minds. They get their degree(s), find work, and perhaps a spouse, and become permanent residents, perhaps citizens. Others may return for other reasons. With jet travel and the internet it is much easier to maintain a connection with a home country.

In 2016, domestic migration into Texas was 532K, and out of Texas was 444K, but the net was only 88K. That is, the net flow was less than 1/10 of the total number of border crossers.

From the 2016 ACS, roughly 2 million persons were living in the USA, who had resided abroad a year earlier. But the Census Bureau estimates net international migration at half of that.

Mumbly. mumbly, hand wavy, the Census Bureau through 2017 reduced the net international flow from 7.2M to 6.7M. If this is due to increased estimate of out migration, this might only require an increase in the percentage of outflow by less than 10%.

The question is then why is this so concentrated in New York, New Jersey, and California.

Speculation,

New York and California by far have the most diverse immigrants. Perhaps  more people have heard of NYC or Los Angeles or San Francisco and move there as part of some fantasy, thinking they can get a job in movies or the theater, etc. It may be more a multi-year visit, than a lifetime migration.

[jimrtex]

Where are these thoughts of NY losing two seats coming from? It seems NY has been on track to lose 1 for 7 years or so. That seat itself has always been likely to be NY22/NY24  as Dems carve up Katko and protect Brindisi or another Dem. I mean of course Dems are going to get messy with the lines in Long Island, Staten Island, and the Hudson Valley, but that is to be expected.

With the addition of the 2018 census, the map has changed.

New York has had a negative second derivative for all 8 years of the decade. It reached a peak in 2015. So not only is it declining, it is declining faster. It is like it is in a dive, and the pilot has pushed the stick forward, and Captain Cuomo is no Captain "Sully".

I think EDS uses a weighted projection. But I was unable to reproduce their projection of the 2020 population, even that based on 2000-2008 estimates.

I did calculate the quotients for the states in the bubble based on the most recent estimate, and I think a more reliable projection would show, California, Minnesota, and New York in contention for the 435th seat.

If New York's decline continues to accelerate it might fall below the others. Minnesota is increasing slightly, while California is slowing down. For 2010-2014 California was growing faster than the US as a whole, while for 2014-2018 it was slower.

By 2020, Florida(+2) and Texas(+3) will have registered their gains, while Alabama and Ohio will have registered their losses.

Illinois and Virginia will just be coming into the picture, but won't see a change before the Census.

[jimrtex]

Is there any set date for the 2018 county estimates?

The scheduled release is April 2019.

[jimrtex]

Naive question: which states (outside at-large) wouldn't redistrict if their house seat total was unaffected in 2020 ?

Every state is required to redistrict for OMOV. Even states like Maine will shuffle a few towns around the edges, as in 2010. It is deeply unlikely that any state would have population changes so uniform that no tweaks to the maps were required.

Estimates for the two New Hampshire districts from the 2017 ACS are that they are about 0.27% of the ideal popoulation. The MOE is about 3 times the deviation, meaning that it is quite possible that NH-2 is the larger rather than smaller district.

I doubt that there is the political will to put Manchester and Nashua in the same district.

It is quite possible that swapping no single town will make the districts more equal, but I suspect that there is some swap of two or three towns that would be. In any case, NH would make an affirmative decision to re-enact the existing districts, rather than not redistricting.

[jimrtex]

This is a work in progress in the beta testing stage, but here's an interactive map of the percentage/numerical change in total population by census tract. There are submaps for 2010-"15", 2000-"15" and 2000-10 (click on Tot Est % or Tot Est #, then the year).

The "2015" data is from the 2013-17 American Community Survey (2015 is the midpoint of that data); the 2000 estimates were based on percentage land area from the 2000-10 tract relationship tables.

https://cinycmaps.com/index.php/population-change/tract-population-change

Red is an increase; blue is a decrease.

Let me know if you encounter any errors or something that looks just plain wrong. Like I said, this is in beta.

Jimrtex - would using percentage population from the relationship tables yield a more accurate result? I'm ultimately going to estimate the changes in racial data, too.

I would look into graying out low-population census tracts - or possibly any of the form

98xxxx (decimal suppressed).

In Texas, most of the dark blue tracts are of this type, and include tracts with airports or military bases outside base housing. The ACS might not even sample these areas. For the census, hotels are asked to report any persons who are resident - it might even be all guests are asked whether they are reported elsewhere, so as to prevent double counting.
I think that for some persons, "usual residence" means where they sleep most often.

A salesman may not have a conventional domicile, or may have established a base in a hotel. There may only be a handful of such persons. Trying to locate them by sample during the ACS may be impossible. Similarly, I doubt that an effort is made to locate persons sleeping rough or in cars.

The census bureau permits delineation of census tracts for areas that are largely non-residential (I think 98xx.xx is the code for these). This might have been new for 2010.

If there is a hotel in (on the edge of) a residential census tract, it might not be counted by the ACS, but this will have little impact on overall population of the tract. But a hotel on an airport might represent most if not all of the population counted in the census, and not be included in the ACS.

I don't understand your last question.

[jimrtex]

What minimum population threshold would you suggest? 10? 25? 50? 100?

My other question is this: The 2000-10 Census Relationship file compares the 2000 census tracts to fit into the 2010 a number of ways. One way is by comparing percentage of land area in a given tract, AREALANDPCT00PT. Another is by comparing population, POPPCT00. Which should I be using when guestimating 2000 in the 2010 tracts?

So far in the beta, I’ve been using land area. Does Pop make more sense? Eventually, I’m going to guestimate change in Hispanic and Non-Hispanic White, etc. populations, too. Neither will be perfect there.

So for a census tract that had a simple division,

POPPCT00 would show the percentage of the 2010 population that was in a particular 2000 Census Tract???

If a 2000 tract was divided, the 2010 tracts might have similar population since the goal would be to create tracts near the target population of 4000. But that might not reflect the situation in 2000. In an area of rapid development one of the 2010 tracts might have been largely built out by 2000, and the other was newly built.

But using area might be even worse. The more developed area may be ready to be placed in an essentially permanent tract. There is no more space that can be developed, and the newly developed areas also include undeveloped land from when the area was part of a rural census tract.

But there are census tracts of 10,000 that could have or should have been divided in 2010, that only have 12,000 persons in 2015, and if divided for 2020 would to some degree  match the 2010 distribution.

The closer the 2000 population of an old 2000 tract matches the combined 2010 population of the new for 2010 tracts, the better the distribution is going to be.

Probably more work than you want to do, but you could use 2000 census blocks with 2010 tract boundaries, but the census blocks have been renumbered, and might not match the 2010 tract boundaries.

[jimrtex]

I don't know if there is data anywhere that incorporates all the revised counts from this.

I found Census' 2010 errata notes.The reason for the errors in Norfolk (and San Diego, Groton, CT, Portsmouth, NH, Pascagoula, MS & Everett, WA) was... a systematic mistake in determining the placement of Navy vessels.

I'll put a fix on my to-do list.

This is another instance of uncertainty/ambiguity of residence for census purposes.

Even if the ships had been located correctly, it would not reflect the domicile of those assigned to the ship.

[jimrtex]

The press embargo begins at 10:00 AM EST December 30. T"he public release should be at 12:01 AM on January 1,

Has CSPAN ever covered a New Year's Eve event?

"Its a virtual sea of pocket protectors here as the revelers await the release of the 2019 state population estimates. Lots of cans of Red Bull and Jolt being opened."

(to man in the crowd) "What do you think of this event?"
(man) "I anticipate that California will lose a representative"
(producer) "Let's stay right here. We can cut back to Times Square at 10 after midnight.

[jimrtex]

There are eight states that are in a band of about 1% of the ideal break point.

Above the line:

NY(26) 1.004
TX(39) 1.003
FL(29) 1.002
MT(2) 1.001
----------------
AL(6) 0.998
MN(7) 0.997
CA (52) 0.995
OH(15) 0.994

At this point, what is more significant is how accurate the estimates are, than the growth trend for the final 9 months of the census period. Overall drops in growth may mean estimates are more accurate. If a state grew at an estimated 30% during a decade, it may be difficult to determine whether it was 29% or 31%. But it is likely an estimated rate of 2.5% is more accurate.

Based on 2009 estimates, Florida and Texas have not yet gained their 39th and 29th seat respectively, while Alabama and Ohio have not lost their 7th and 16th seat respectively.

New York year-to-year change has dropped every year this decade. That is, its second derivative is negative, and its population is in parabolic decline. After the 2011 estimate, it was projected that New York would keep its 27th seat. Since then, smaller increases, which have have become decreases since 2016 put New York at risk of losing two representatives. A projection based on estimates from 2010-2019 would result in a surplus of around 130K. But the average annual gain from 2010 to 2019 was around 6K, while the loss for 2018-2019 was 76K. If we project a loss of 95K for 2020 based on recent trends, that would be a deficit of 100K (75K for nine months) which would eat into more than half of the projected surplus. Much will depend on how accurate the revisions in estimates of international migration were. International emigration is difficult to measure since it depends on measuring persons who no longer exist.

Birth records are accurate. Death records are accurate. Estimates of domestic migration based on IRS and SSA records may be reasonably accurate. If New York keeps a 26th representative it may be lost before the 2020 census results are announced. New York is projected to lose 1.634 representatives (from 27.243 to 25.609). They are in a favorable position for rounding.

Texas year-to-year peaked near mid-decade with $100/bbl oil, and then dropped when the price drop to $30/bbl. By 2019, the price has been around $55/bbl which is profitable in the Permian Basin, but not necessarily elsewhere. The year-to-year increase has recovered to near the decade average, and the Texas increase as a share of the national increase was a decade high of 23.7%. Texas has a 147K surplus for the 39th seat and should maintain that. The caveat is whether the Census estimate of 16.6% (2nd highest to Utah) has been accurate.

Florida started the decade recovering from the housing bubble, but by mid-decade growth in 2016 was almost double that of 2017. Since then growth has dropped rather sharply, with 2019 growth almost back to to 2011 levels. Continuation of the trend would make 2020 less than 2011, eating into Florida's 106K surplus. Florida may benefit from the short 9-month interval to the Census. Again the final determination will largely hinge on the quality of estimates.

Montana has been just below the population necessary for a second seat ever since it lost its second seat. Some years it looked might it gain it back and others not. With an economy highly dependent on natural resources and tourism, growth can be volatile. Population is projected to increase by 8.7%, slightly above the national average of 6.8%, but enough for a gain of 0.023 representatives to put Montana over the cusp. With a surplus of just 3300, Montana may be harmed by the April 1 Census date, as seasonal Montanans may not have returned from Arizona.

Alabama had its peak increase in 2019, but it is projected to be about 15K short of keeping the 7th seat. But its total growth for the decade is only 133K. It would essentially have to double its growth rate of the last decade, or hope that some people have been missed.

Minnesota may illustrate an apportionment paradox. It is projected to increase 6.86% for the decade, versus 6.83% for the US as a whole. It's share of the US population is increasing, while its share of representation is decreasing. The entitlement is an increase from 7.472 to 7.474. In effect, Minnesota was lucky not to have lost its 8th seat in 2010.
Minnesota is projected to be about 23K short of keeping its 8th seat. But that it is about 2/3 of its average annual growth. 2019 was Minnesota's 3rd lowest year. It is not likely to have that much in a 9-month period, so its hope depends on inaccurate estimates over time.

California over the first half of the decade had strong growth of over 300K per year. Since 2015 it has plummeted (249K, 191K, 103K, 51K). Based on recent trends, California may have essentially no population growth this year. California may be experiencing a delayed effect of the recovery from the housing bubble. Home owners who were upside down on mortgages have been able to finally cash out, and move to states with more affordable housing. California is 222K short of keeping a 53rd representative and may pile up as much more in the final 9 months of the decade.

Ohio is 80K short of keeping its 16th seat. It's growth has been meager and sporadic. 80K is equivalent to an average of 5 years growth, and 3 of the best year.

[jimrtex]

2020 Pop is projected population in millions for April 2020, based on April 2010 (base) and July 2019 estimate.

2010 Ent is based on geometric mean.  sqrt((pop/quota)² + 0.25). Quota is adjusted so that states sum to 435 (about 0.43%). A state with n.500 will be near the dividing line for gaining a n+1 representative.

2020 Ent is based on geometric mean.

2010-2020 is difference in 2020 entitlement minus 2010 entitlement. For example, Mississippi can be said to have lost about 1/4 a representative.

Rank Val is the ranking quotient, for a state receiving an n+1 district if 2020 entitlement was n.xxxx. pop/sqrt(n*(n+1)). Values are normalized by dividing by the adjusted quota, so that values above 1 will generally indicate that the state will gain the extra seat (since 435 seats are apportioned, it is possible that some states above 1 will not gain a seat, or that some below 1 will gain a seat. In this case it happens that Montana(2) with 1.001 is the 435th seat.

2020 is the projected apportionment for 2020.

Change is change in apportionment from 2010.

Next is guesstimate of next change based on 2010-2020 trend. If blank, no change is expected through 2050. Trends out that far are unreliable anyway.

State	2020 Pop	2010 Ent	2020 Ent	2010-2020	Rank Val	2020	Change	Next
Hawaii	1.420	1.976	1.935	-0.042	1.322	2	0
New Hampshire	1.363	1.917	1.862	-0.055	1.268	2	0
Maine	1.346	1.933	1.840	-0.093	1.252	2	0
Mississippi	2.977	4.202	3.948	-0.253	1.131	4	0	L 2040/50
New Mexico	2.100	2.937	2.808	-0.130	1.128	3	0	L 2040/50
Kansas	2.918	4.041	3.872	-0.169	1.108	4	0	L 2040/50
Kentucky	4.478	6.120	5.913	-0.207	1.076	6	0	L 2040/50
Maryland	6.068	8.131	8.000	-0.131	1.067	8	0
South Carolina	5.194	6.521	6.852	0.331	1.054	7	0	G 2040/50
Connecticut	3.565	5.049	4.717	-0.332	1.049	5	0	L 2030/40
Indiana	6.753	9.128	8.899	-0.229	1.047	9	0	L 2040/50
Nebraska	1.943	2.615	2.605	-0.010	1.044	3	0
North Carolina	10.569	13.413	13.915	0.502	1.031	14	+1	G 2030/40
Wisconsin	5.834	8.010	7.692	-0.318	1.026	8	0	L2030
Colorado	5.822	7.087	7.677	0.590	1.024	8	+1	G 2040
Pennsylvania	12.810	17.862	16.862	-1.000	1.022	17	-1	L 2030
Oregon	4.251	5.408	5.615	0.207	1.021	6	+1
Arizona	7.356	8.999	9.691	0.692	1.020	10	+1	G 2040
New Jersey	8.890	12.368	11.707	-0.661	1.018	12	0	L 2030
Illinois	12.659	18.043	16.663	-1.380	1.010	17	-1	LL 2030
New York	19.460	27.243	25.609	-1.634	1.004	26	-1	LL 2030
Texas	29.333	35.349	38.597	3.248	1.003	39	+3	GGG 2030
Florida	21.710	26.437	28.570	2.133	1.002	29	+2	GG 2030
Montana	1.075	1.478	1.501	0.023	1.001	2	+1
Alabama	4.913	6.738	6.484	-0.254	0.998	6	-1
Minnesota	5.668	7.472	7.474	0.002	0.997	7	-1
California	39.701	52.368	52.238	-0.130	0.995	52	-1
Ohio	11.702	16.224	15.404	-0.820	0.994	15	-1	L 2040
Rhode Island	1.060	1.562	1.481	-0.081	0.986	1	-1
Virginia	8.580	11.258	11.301	0.043	0.983	11	0
Michigan	9.995	13.902	13.161	-0.742	0.975	13	-1	L 2030
Georgia	10.697	13.628	14.083	0.455	0.971	14	0	G 2030/40
Idaho	1.806	2.260	2.428	0.169	0.970	2	0	G 2030
Washington	7.692	9.465	10.133	0.668	0.965	10	0	G 2030
West Virginia	1.787	2.652	2.404	-0.248	0.960	2	-1
Massachusetts	6.921	9.217	9.120	-0.097	0.960	9	0
Oklahoma	3.974	5.297	5.253	-0.044	0.955	5	0
Utah	3.245	3.917	4.298	0.381	0.955	4	0	G 2030
Missouri	6.150	8.433	8.107	-0.326	0.954	8	0	L 2040/50
Tennessee	6.870	8.934	9.053	0.118	0.953	9	0
Louisiana	4.658	6.392	6.149	-0.243	0.946	6	0	L 2050
Iowa	3.164	4.312	4.193	-0.119	0.931	4	0
Nevada	3.113	3.829	4.126	0.298	0.916	4	0	G 2030/40
Delaware	0.980	1.358	1.383	0.026	0.912	1	0
Arkansas	3.026	4.129	4.013	-0.116	0.890	4	0
South Dakota	0.891	1.249	1.274	0.025	0.829	1	0
North Dakota	0.770	1.069	1.130	0.060	0.716	1	0
Alaska	0.733	1.117	1.087	-0.030	0.682	1	0
Vermont	0.624	1.012	0.961	-0.051	0.580	1	0
Wyoming	0.580	0.937	0.912	-0.025	0.540	1	0

[jimrtex]

Good breakdown jimrtex. The only thing that I would add is that CA is pouring in millions to assist the census 'discover' residents. The South Valley has a lot of hard-to-reach migratory groups, and CA is pouring in money to ensure every resident is counted, especially those groups missed in 2010. The margin of error around projections therefore is larger here and in other border states, so it's one thing working in the states favor. In contrast, TX has a lot of hard-to-reach groups as well, but the state is putting aside 0$ to assist in contacting their residents (for political reasons of course) so the MOE around those projections is less in their favor.

It is unclear whether you are referring to the San Joaquin Valley, or southern San Fernando Valley.

Estimates of undercount(PDF)

The undercount for California in 2020 is estimated to be 0.60%, while California is estimated to be short by 0.5% of holding on to its 53rd seat.

There might be areas where the undercount was higher, but also lower such as in Marin. And is the mass outreach effective?

[jimrtex]

Based on 2010-2020 change Texas will surpass California in population in late 2053, when both states will have over 49 million.

[jimrtex]

Has California ever lost a seat before?

No. 2010 is the first apportionment it did not gain a seat.

From 1850 to 2010:

2, 3, 4, 6, 7, 8, 11, -, 20, 23, 30, 38, 43, 45, 52, 53, 53

Current streaks:

FL 11* (since 1890)
TX 7 (since 1940)
AZ 6 (since 1950)

All are expected to extend the streak.

*If there had been an apportionment in 1920, Florida's streak would have broken. It only gained one seat between 1910 and 1930. Thus it would have had the same in 1910 and 1920, or in 1920 and 1940.

Current losing streaks.

PA 9 (since 1910, it lost two between 1910 and 1930, but it is possible those were both between 1920 and 1930).
NY 7 (since 1940)
OH 5 (since 1950)
MI 4 (since 1960)
IL 4 (since 1960)

Other long winning streaks:

CA 14 (1850 to 2000)
IL 9 (1820-1910)
MO 8 (1820-1900)
TX 7 (1850-1930)
NY 5 (1789*-1830)
PA 5 (1789*-1830)
PA 5 (1860-1910)
WI 5 (1850-1900)

1789* apportionment under Constitution, NY and PA gained under 1790 Census.

Other long losing streaks:

VA 6 (1810-1870*)

*Loss in 1870 due to separation of WV.

States that have never lost representatives:

AK, AZ, CA, CO, FL, HI, ID, NV, NM, OR, TX, UT, WA, WY.

[jimrtex]

Does anyone know when in May the 2019 estimates at the municipal level are to come out?  Or will that even be in May as planned or will it be delayed due to COVID-19?  The March release of county-level estimates wasn't delayed, but those were likely pretty much finalized by the time things started getting crazy.

12:01 AM EDT, 21 May 2020.

[jimrtex]

Wonder what it takes to qualify as "media" these days.  If you were a representative of the Talk Elections Blog, would that qualify?

Guidelines for Embargo Access to Media Embargo Site

The key would be the primary purpose of the website is dissemination of the news. Research and publication of election results might be considered news. Conceivably if Dave Leip actively participated with presenting two bylined stories per month, he might qualify.

Embargo access is granted to individuals, with the condition that they not share the data with colleagues. This would like disqualify anyone here.

[jimrtex]

The Northeast region as a whole lost population last year.    10 states lost population too,  one higher than last year. 

New York lost almost 77k people, does that mean it'll lose two congressional districts?

Its looking extremely likely at this point.

And one could make the argument that both seats should come at the expense of Upstate New York.

The math doesn’t add up for two upstate losses. The more likely result is that one seat upstate is lost along with one Long Island/East NYC seat.

The math might not add up to 2 lost seats at all. NY-26 is literally on the bubble in many of the analyses I've seen.

I'd bet that it will lose 2, though.

Cinyc, what odds do you place on NY losing 2 seats rather than 1? Whether NY loses two CD's or one has a big impact on the options for the border of the current NY-19 given NY's funnel shape.  NY clearly has the oddest shape of all 50 states in my opinion.

How accurate is the census when it asks where one lived on April 1 four months later? Is there any data that addresses that issue? I assume there will be more late responses this year than in the past.

This is the analysis I made last December:

New York year-to-year change has dropped every year this decade. That is, its second derivative is negative, and its population is in parabolic decline. After the 2011 estimate, it was projected that New York would keep its 27th seat. Since then, smaller increases, which have have become decreases since 2016 put New York at risk of losing two representatives. A projection based on estimates from 2010-2019 would result in a surplus of around 130K. But the average annual gain from 2010 to 2019 was around 6K, while the loss for 2018-2019 was 76K. If we project a loss of 95K for 2020 based on recent trends, that would be a deficit of 100K (75K for nine months) which would eat into more than half of the projected surplus. Much will depend on how accurate the revisions in estimates of international migration were. International emigration is difficult to measure since it depends on measuring persons who no longer exist.

Birth records are accurate. Death records are accurate. Estimates of domestic migration based on IRS and SSA records may be reasonably accurate. If New York keeps a 26th representative it may be lost before the 2020 census results are announced. New York is projected to lose 1.634 representatives (from 27.243 to 25.609). They are in a favorable position for rounding.

Our estimates of seats lost or gained are based on projections of the census estimates. Assuming that census estimates are accurate, our projections become more accurate as the decade wears on because we are projecting over a shorter period of time, in this case for the 9 months between July 2019 and April 2020.

But our projections are simple-minded. We assume that the trend from 2010-2019 will continue. Or we assume the trend from 2017-2019 will continue. But in New York, the estimated loss from 2018-2019 was greater than for 2017-2018, and in the early part of the decade, New York was gaining. Cities in general were gaining population post recession, as people were having to defer retirement or a move to the suburbs. Some were having to remain at their parents house, or stay coupled, even if not married.

The census estimates have to account for emigration. While immigration to NYC is very high, emigration is also high. Someone who is overstaying their visa might be included in an estimate. But if they move back to Tajikistan will it be noticed? Refinements of census estimates had a particular effect on New York. It has lots of immigrants from lots of places,  but they may return to those places as well. The effect was smaller in California and Texas because immigrants from Latin America are less likely to return.

And finally, apportionment of representatives is among the states. For New York to lose a second seat, another state has to gain it. There are eight states vying for the final four seats.

NY(26) 1.004
TX(39) 1.003
FL(29) 1.002
MT(2) 1.001
----------------
AL(6) 0.998
MN(7) 0.997
CA (52) 0.995
OH(15) 0.994

New York has been faltering as has California. Texas has been OK, not as strong as earlier but not slowing to a halt. Florida has been strengthening since recovering from the housing bubble, and is gaining Puerto Ricans, as well as Venezuelans who can escape.

That leaves perhaps New York, Montana, Alabama, and Minnesota competing for the last two. None are really grabbing those last two.

As you probably know, half the New York population is insular: Long, Staten, and Manhattan; and half on the mainland Bronx, Westchester and points north and west. If there an odd number of districts, then you will have a district spanning from either Manhattan or Queens to the Bronx, with 5-1/2 districts south of the 19th. If there is an even number, then there will be 5 districts south of the 19th, which will yank the district south.

Assuming that the new districts will be drawn by a federal judge, or the redistricting commission produces something reasonable, NY-23 is the odd man out.

You aren't going to push districts out of Buffalo, Rochester, Syracuse, Albany or the North Country. NY-23 and NY-27 might seem to be available but that can only be done with adding fingers into their territory.

So NY-27 slides east taking Binghamton, perhaps losing Chautauqua. It will be quite vulnerable in 2030 as the four districts to the north will need to expand. NY-24 shifts eastward while remaining anchored in Syracuse (Wayne is a better fit with Rochester anyway). NY-21 will take Utica, becoming less and less a North Country district, but still identified as such. NY-19 may push a bit west.

Delays due to COVID-19 have pushed the announcement of congressional apportionment to March 2021, and release of the block-level number to June 2021 (this is not official because the deadlines are set in statute, but there is zero chance of this not happening).

In New York, there is also the prison-adjustment. This will make it almost impossible to re-ward Hudson before the July 2021 primary. Perhaps partisan primaries can be eliminated and go with a non-partisan general election in November, with a runoff if necessary.

[jimrtex]

https://www.mercurynews.com/2020/07/08/study-in-a-first-california-poised-to-lose-house-seats/?utm_source=twitter.com&utm_campaign=socialflow&utm_medium=social&utm_content=tw-mercnews

It smells like clickbait, but at the same time California has a large number of 'hard to count' census regions, mostly areas in the valley with Rural Hispanic populations. The article mostly talks about the effects and why it might happen, corona and ensuing response rates, rather than the math behind it - typical of puff pieces. So, could our resident calculators like Jimrtex and Cinyc voice in on the scenario outlined about California losing two seats.

It was questionable whether California would lose one seat, but that was largely confirmed if you took into account the declining rate of growth toward the end of the decade.

Here is the news release that the article is based on.

Southern California Faces Loss of One Congressional District, May narrowly hold on to second

Note that like the news article bungles Southern California and California together.

This may be the report underlying the news release. Note if you click on Download you can get the report in PDF format.

Winners and Losers: The 2020 Census and California's 2021 Redistricting

The underlying estimates are based on the 2017 one-year estimates projected forward to 2020. They make a small error by not accounting for estimates being based on July 1, and the census occurring on April 1. The ACS is a year-round collection, with July 1 being midpoint of the year.

A better projection would use the 2019 county estimates dividing by 9.25 and projecting forward by adding 0.75 per year.

They divide Los Angeles into 5 regions. You could make projections based on the 2018 ACS, project forward to 2020, and then control the regional populations to the countywide projection.

Their estimates for congressional representation take into account a loss of one representative for California, but due to rounding to 0.1 districts, the 2010 counts total 52.8, leaving the loss of 0.2 districts unaccounted for.

The only ESRI report suggesting a loss of another seat that I could find was based on a supposition of 4 seats being apportioned to a state of Puerto Rico.

Apportionment for California is kind of weird. It gains/loses a seat for about every 2% gain or loss relative to the US (1/52 = 1.92%). But apportionment is among all 50 states, so while the quotients for California are evenly spaced, that of the collective 50 states is not. So on the priority list California might be 435th, 431st, and 419th, rather than a regular spacing of every 8th or 9th seat.

The story is naive about losing a particular district. Imagine that if in 2010, California had been informed they only had 52 seats, but had already completed their 53-district plan.

Every district would have to gain about 15,000 persons. So you start working south from the Golden Gate bridge, and Oregon and converging around Stockton. The first district nibbles 15,000 from a neighbor, which takes 30,000 from its neighbor and so on. By the time you get to Bakersfield, Palmdale, Apple Valley, you are shifting half a district around, pairing incumbents, and creating open seats.